arc agi
30 articles about arc agi in AI news
Prime Intellect's Prime Agent Hits 95.5% on ARC-AGI-3 With Opus 5
Prime Intellect's open-source Prime Agent scored 95.5% on ARC-AGI-3 with Opus 5, exceeding the human baseline via a self-improving RLM harness.
OpenAI hits 38.3% on ARC-AGI-3 with custom API, bypassing official harness
OpenAI's GPT-5.6 Sol scored 38.3% on ARC-AGI-3 with custom API settings, beating Opus 5's 30.2%, but scored 7.8% in the official harness, exposing benchmark parity issues.
DeepSeek V3.2 Agent Hits 67% on ARC-AGI-1 Without Fine-Tuning
Moghe & Chin achieve 67.25% pass@2 on ARC-AGI-1 using DeepSeek V3.2 in non-thinking mode at $0.62/task, with no fine-tuning. The work demonstrates agent architecture alone can lift a 15.50% baseline by ~52 points.
AI Researcher Kimmonismus Predicts AGI Within 6-12 Months, Widespread Worker Replacement in 1-2 Years
Independent AI researcher Kimmonismus predicts AGI will arrive within 6-12 months, with widespread worker displacement following in 1-2 years. The forecast, shared on X, adds to a growing chorus of near-term AGI predictions from industry figures.
Symbolica's Agentica SDK Scores 36.08% on ARC-AGI-3, Claiming Cost-Effective Agentic Breakthrough
Symbolica's Agentica SDK reportedly achieved a 36.08% score on the new ARC-AGI-3 benchmark in one day, using an agentic approach claimed to be far cheaper than brute-forcing with a frontier model.
Frontier AI Models Reportedly Score Below 1% on ARC-AGI v3 Benchmark
A social media post claims frontier AI models have achieved below 1% performance on the ARC-AGI v3 benchmark, suggesting a potential saturation point for current scaling approaches. No specific models or scores were disclosed.
Claude Code's Six-Layer Architecture: Harness, Not Magic
Claude Code's six-layer architecture uses a 3-layer context compressor at 92% threshold and Redis-based multi-agent FSM protocol. The model is just one node in a harness.
The Fragility of China's Open-Source AI: New Research Reveals Capability Gaps
New empirical evidence reveals Chinese open-weight AI models show significant fragility compared to frontier closed models, excelling in narrow domains but struggling with general tasks and out-of-distribution challenges.
AgiBot WITA-Omni Scores 85.21 on DailyOmni, Beats Gemini
AgiBot WITA-Omni scores 85.21 on DailyOmni benchmark, beating Google Gemini, ByteDance Doubao, and Alibaba Qwen with a novel Thinker-Talker-Actor architecture.
How a Healthcare Startup Used Claude Code to Ship 66 Architecture Tickets in 4 Hours
Claude Code can autonomously execute complex architecture work when given proper domain expertise, ticket planning, and execution authority—no magic required.
Roman Yampolskiy: 'AGI is a Question of Cost, Not Time' as Scaling Laws Hold
AI safety researcher Roman Yampolskiy argues that achieving AGI is now a matter of computational and financial resources, not theoretical possibility, citing the continued validity of scaling laws and early signs of recursive self-improvement.
The Threshold of Weak AGI: How Modern AI Systems Are Quietly Passing Historic Milestones
Leading AI researcher Ethan Mollick highlights that current models like GPT-4.5 have already achieved several key benchmarks for 'weak AGI,' including Turing Test equivalents and complex reasoning tasks, with only one remaining historical challenge.
Spine Swarms: How an 8-Person Team Outperformed AI Giants in Deep Research
A small team of engineers has developed Spine Swarms, an AI system that reportedly outperforms Google, Perplexity, Claude, and GPT-5.2 in deep research tasks. This breakthrough demonstrates how agile teams can compete with tech giants in specialized AI applications.
The Hidden Achilles' Heel of AI Imaging: How Tiny Mismatches Cripple Compressive Vision Systems
New research reveals that state-of-the-art AI for compressive imaging catastrophically fails when its mathematical assumptions about hardware don't match reality. The InverseNet benchmark shows performance drops of 10-21 dB, eliminating AI's advantage over classical methods in real-world deployment.
Neural Paging: The Memory Management Breakthrough for Next-Gen AI Agents
Researchers propose Neural Paging, a hierarchical architecture that decouples symbolic reasoning from information management in AI agents. This approach dramatically reduces computational complexity for long-horizon reasoning tasks, moving from quadratic to linear scaling with context window size.
The Unlearning Illusion: New Research Exposes Critical Flaws in AI Memory Removal
Researchers reveal that current methods for making AI models 'forget' information are surprisingly fragile. A new dynamic testing framework shows that simple query modifications can recover supposedly erased knowledge, exposing significant safety and compliance risks.
Google DeepMind Unveils 'Intelligent AI Delegates': A Paradigm Shift in Autonomous Agent Architecture
Google DeepMind has introduced a groundbreaking framework called 'Intelligent AI Delegates' that fundamentally reimagines how AI agents operate. This new architecture enables more autonomous, efficient, and collaborative problem-solving by allowing AI systems to delegate tasks dynamically.
WWD Archive Traces Calvin Klein's Off-Court Tennis Legacy
WWD's archive feature traces Calvin Klein's off-court tennis style across three decades, using its own reporting. The piece connects 1980s minimalism to today's tennis-fashion boom.
Claude Code's Architecture Explained
Master Claude Code's 6-layer architecture—harness, context window, subagents, MCP, skills, hooks, CLAUDE.md—to optimize token usage, parallelism, and agent reliability. First sentence: Claude Code's 6 layers (harness, context window, subagents, MCP, skills, hooks, CLAUDE.md) determine your setup's success.
Building a Multimodal Vector Search Platform for Product Catalogs
Insider Engineering shares practical lessons from building a multimodal vector search platform for product catalogs, covering multitenancy, GPU economics, and infrastructure surprises. The post provides actionable insights for retail AI teams considering similar systems.
SemiAnalysis Launches Mythos AI Research Platform
SemiAnalysis launched Mythos, a proprietary AI research platform for semiconductor and AI industry analysis, announced via Twitter on March 5, 2026.
111-Page Survey Maps 5 AGI Levels: Responder to Ecosystem
111-page survey from US/China labs defines 5 AGI levels, argues epistemic exploration — not better answering — is key. Challenges scaling orthodoxy.
Hassabis: AGI by 2030 Is 'Singularity-Level' Shift, Society Unprepared
Demis Hassabis warned AGI around 2030 will be a singularity-level event. He says society has little time to prepare for a revolution ten times faster than the Industrial Revolution.
AgingBench: AI Agents Lose Reliability Over Time & Memory Fails
UT Austin paper finds AI agents degrade over time via memory errors. Proposes AgingBench to measure reliability decay across sessions.
Dario Amodei Predicts AGI by 2028, Cites 'Mythos' Step Change
Dario Amodei predicts AGI by 2028, citing a step-function advance in 2026. He envisions millions of autonomous agents in datacenters.
Demis Hassabis: AGI Components Exist, Missing Continual Learning
Demis Hassabis claimed AGI components exist but continual learning and memory remain unsolved. The statement reframes the AGI debate from foundational to incremental.
Large Memory Models: New Architecture Beyond RAG and Vector Search
Researchers with 160+ Nature and ICLR publications have built Large Memory Models (LMMs), a new architecture designed to emulate human memory processes, offering an alternative to RAG and vector search paradigms.
OpenAI Agents Now Ask Questions Good Enough for Research Papers
Sébastien Bubeck revealed on the OpenAI Podcast that internal AI agents now ask research questions so insightful they're inspiring papers and correcting published mistakes, with a 1-2 year timeline for full researcher-level capabilities.
New Research Models 'Exploration Saturation' in Recommender Systems
A research paper analyzes 'exploration saturation'—the point where more diverse recommendations hurt user utility. Findings show this saturation point is user-dependent, challenging the standard practice of applying uniform fairness or novelty pressure across all users.
Researchers Achieve Ultra-Long-Horizon Agentic Science with Cohesive AI Agents
A research team has developed AI agents capable of executing and maintaining coherent, long-horizon scientific research workflows. This addresses a core challenge in creating autonomous systems for complex discovery.